International Geology Review, Vol. 48, 2006, p. 479–493.Copyright © 2006 by V. H. Winston & Son, Inc. All rights reserved.

Simple Kinematics of Subduction ZonesC. DOGLIONI,1 E. CARMINATI, AND M. CUFFARO

Dipartimento di Scienze della Terra, Università La Sapienza, P. Le A. Moro 5, Box 11, 00185 Rome, Italy

Abstract

Two main types of subduction zones can be distinguished: (1) those where the subduction hingemigrates away from the upper plate; and (2) those in which the subduction hinge migrates toward theupper plate. Apart from a few exceptions, this distinction seems to apply particularly for W-directedsubduction zones and E- or NE-directed subduction zones, respectively. Moreover, the rate of sub-duction is larger than the convergence rate along W-directed subduction zones, whereas it is smalleralong E- or NE-directed subduction zones. Along W-inclined slabs, the subduction rate is theconvergence rate plus the slab retreat rate, which tends to equal the backarc extension rate. AlongE- or NE-dipping slabs, the shortening in the upper plate decreases the subduction rate, and notypical backarc basin forms.

Relative to the mantle, the W-directed slab hinges are fixed, whereas they move west or south-west along E- or NE-directed subduction zones. The single vergent accretionary prism related toW-directed subduction zones is formed at the expenses of the shallow layers of the lower plate. Theorogens generated by E- and NE-dipping subductions are double vergent since the early stages, andupper plate rocks mostly compose them until continental collision takes place, eventually involvingthe rocks of the lower plate more extensively. The convergence/shortening ratio in this type oforogens is higher than 1 and is inversely proportional to the viscosity of the upper plate. The higherthe viscosity, the smaller the shortening and faster the subduction rate. The convergence/shorteningratio along W-directed subduction zones is instead generally lower than 1.

W-directed subduction zones provide larger volumes of lithospheric return into the mantle thanthe opposite E- or NE-directed subduction zones. The latter may contribute to refertilize the shallowupper mantle. These observations suggest that subduction zones are passive features with respect tothe far-field velocity of plate motions and the relative “eastward” mantle flow, implicit with a net“westward” rotation of the lithosphere.

Introduction

THE SUBDUCTION PROCESS is a vital ingredient ofplate tectonics (Anderson, 1989, 2001; Stern, 2002;Conrad and Lithgow-Bertelloni, 2003). Most of theoceanic lithosphere is recycled into the mantle, butalso some continental lithosphere can subduct(Ampferer, 1906; Ranalli et al., 2000; Hermann,2002). There still are doubts on the initiation of thesubduction process (Spence, 1987; Niu et al., 2003),but once activated, it may work for several tens ofmillion years. Subduction zones have a relatedaccretionary prism or orogen, a subduction hingethat can migrate (Garfunkel et al., 1986), and a largevariety of geophysical, magmatological and geologi-cal signatures (Isacks and Molnar, 1971; Jarrard,1986; Royden and Burchfiel, 1989; Tatsumi andEggins, 1995; Hacker et al., 2003; Carminati et al.,

2004; Ernst, 2005). In this article we only discussthe most striking asymmetries among subductionzones and their kinematics.

Subduction zones and related orogens show sig-nificant differences as a function of their polarity.W-directed slabs are generally very steep (up to 90°)and deep, they have a cogenetic backarc basin; therelated single-vergent accretionary prism has lowelevation, it is mostly composed of shallow rocks,and has a frontal deep trench or foredeep (Doglioniet al., 1999). The E- or NE-directed subductionzones are less inclined (30–60°), and seismicitygenerally dies at about 300 km, apart some deeperclusters close to the upper-lower mantle transition(Omori et al., 2004). The related orogens have highmorphological and structural elevation, wide out-crops of basement rocks, and two shallower trenchesor foredeeps at the fronts of the double-vergent belt,i.e., the forebelt and the retrobelt. These differencescan be ascribed to shallower upper-crustal decolle-1Corresponding author; email: carlo.doglioni@uniroma1.it

4790020-6814/06/876/479-15 $25.00

480 DOGLIONI ET AL.

ments typical of W-directed subduction zones withrespect to the deeper lithospheric-rooted decolle-ments occurring along E- and NE-directed subduc-tion zones. The asymmetry can be recognized notsimply comparing W- versus E-directed subductionzones, but subductions along the sinusoidal flow ofabsolute plate motions that undulates from WNW inthe Pacific, E-W in the Atlantic and NE to NNEfrom eastern Africa, the Indian Ocean, and theHimalayas. In the hotspot reference frame a “west-ward” rotation of the lithosphere can be observed(Gripp and Gordon, 2002). The origin of this netrotation of the lithosphere (Bostrom, 1971) is stillunder debate (Scoppola et al., 2006), but it amountsto about 4.9 cm/yr at his equator (Gripp and Gordon,2002). This implies that the plate motion flow issignificantly polarized toward the “west”, i.e.,following or opposing the“eastward” relative mantleflow (Doglioni et al., 1999).

Regional foreland monoclines also show a signa-ture such as steep dip and fast subsidence rates forW-directed subduction zones, and shallow dip andslow subsidence rates for E- or NE-directed subduc-tion zones (Doglioni, 1994; Mariotti and Doglioni,2000). This is a paradox because, along the highestmountains such as the Andes and Himalayas, weshould expect the deepest trenches and foredeeps,but we rather observe these features along the oppo-site subduction zones where there is not a signifi-cant lithostatic load, like the Marianas trench, theApennines, and Carpathians foredeep. The relative“eastward” mantle flow could favor the bending ofthe slab and the superficial subsidence along thehinge of W-directed subduction zones. The mantlecounterflow rather should contrast to the subsidencealong the opposite subduction zones, providing aforce sustaining the slab (Doglioni, 1994). AlthoughW-directed subduction-related prisms have low-grade metamorphism, the E or NE subduction–related orogens at the collisional stage may exhibitUHP rocks (Doglioni et al., 1999). Metallogenesisalso appears to be controlled by subduction style(Mitchell and Garson, 1981). The Mariana typesubduction is characterized by Kuroko or similarvolcanogenic sulfide deposits (Nishiwaki andUyeda, 1983). Porphyry copper deposits are insteadconcentrated in collisional settings and Andean-type subduction zones.

Within the two opposite end-members, a numberof different settings can occur. For example alongthe E-or NE-directed subduction zones, oceanicslabs underlie continental lithosphere such as the

Andes, which may or may not evolve to continent-continent collision such as the Alps or Himalayas(Ernst, 2005). Along W-directed subduction zones,there also are variable compositions of the lowerplate (both oceanic and continental lithosphere,e.g., Marianas and Apennines, Banda arc) and vari-able depth of the basal decollement plane, resultingin variations in the volume of the related accretion-ary prism (e.g., Bigi et al., 2003; Lenci et al., 2004).

Royden and Burchfiel (1989) have proposed thatthe opposite tectonic style between Alps and Apen-nines can be obtained by kinematic scenarios inwhich the slab retreat is slower or faster thanthe convergence rate. Finite element models haveconfirmed this relation (Waschbusch and Beaumont,1996). However, these differences appear to bemore sensitive to the geographic direction of thesubduction (Doglioni et al., 1999). Based on the twoend-member subduction zones, i.e., W- versus E- orNE-directed slabs, the first-order kinematics of theopposite subduction zones are briefly outlined. Inthis paper, we discuss the relation between conver-gence rate and subduction rate, which are differentas a function of the subduction polarity, and thecrucial behavior of the subduction hinge that canmigrate away or toward the upper plate (Heuret andLallemand 2005; Lallemand et al., 2005).

Basic Kinematics of Subduction Zones

Considering a fixed plate in the hanging wall of asubduction zone (U), and the subduction hinge (H)that is a point migrating over the downgoing litho-sphere, we can observe an opposite behavior alongconvergent margins (Fig. 1). Let us assume a fixedpoint in the lower plate (L). The resulting subduc-tion rate (S) can be calculated from the followingsimple relation, S = HL + UL, where HL and UL arethe rates of relative motions, respectively, betweenH and L, and U and L. In this frame, HL = UH – UL.Movements diverging relative to the upper plate areassumed as positive, and converging negative.

As evident from Figure 1, the subduction hingeH can migrate away from point U (upper panel a), ortoward it (lower panel b). In the first case, the UH isnegative, whereas it is positive in the second case.In other words, the subduction hinge can eitherdiverge from the upper plate, or approach it. Thisbipolar behavior is generally observable when com-paring W-directed subduction zones versus E- orNE-directed subduction zones.

KINEMATICS OF SUBDUCTION ZONES 481

Assuming a fixed convergence rate of 8 cm/yr inboth cases, in case the subduction hinge migratesaway from the upper plate at a rate of 2 cm/yr (A),the subduction rate will increase the convergencerate by the amount of slab retreat, i.e., the hingemigration, resulting in 10 cm/yr. If the subductionhinge migrates toward the upper hangingwall plateat –6 cm/yr because part of the convergence isabsorbed by the shortening in the orogen (B), theresulting subduction rate will be smaller, i.e., only2 cm/yr.

The two opposite settings are characterizedrespectively by a much smaller, single-vergentaccretionary prism, and the occurrence of a backarcbasin (A), whereas in the opposite case there formsa double-vergent elevated orogen (B). This differ-

ence in kinematics can be applied to a comparisonof the Apennines and Alps (Carminati et al., 2004),or to the western versus eastern Pacific subductionzones. Therefore, behavior of the subduction hingethat can decrease or increase the subduction rate iscrucial. This shows that the rate of subduction is notequal to the rate of convergence unless the subduc-tion hinge is fixed relative to the upper plate.

Let us assume another frame of reference forsubduction zones—e.g., not relative to the upperplate, but with respect to the mantle. In the exampleof a W-directed subduction (Fig. 2), the upper plateU moves westward at 2 cm/yr. The hinge H is fixedrelative to the mantle, but it is a transient point,shifting on the downgoing plate at the mantle veloc-ity. The lower plate L moves westward at a speed of11 cm/yr and the convergence rate between upperand lower plates is then 9 cm/yr. The backarcspreading, which is the rate of motion of U relativeto H, is 2 cm/yr, and the resulting subduction rate is

FIG. 1. Basic kinematics of subduction zones, assuming afixed upper plate U. Movements diverging relative to theupper plate are positive, whereas they are negative when con-verging. A. The trench diverges from the upper plate, and thesubduction rate is faster than the convergence rate. In thiscase, the backarc extension increases the convergence rate. B.The trench converges toward the upper plate and the resultingsubduction rate is slower with respect to the convergence rate.That is because the shortening in the orogen decreases thesubduction rate.

FIG. 2. Absolute plate motions of a W-directed subductionzone. The slab is anchored in the mantle, and the subductionrate is faster than the convergence rate. The subduction hingeH is fixed relative to the mantle, but is moving east relative tothe upper plate. Therefore in both W- and E-/NE-directedsubduction zones, the hinge migrates eastward relative to theupper plate.

482 DOGLIONI ET AL.

11 cm/yr. The subduction rate is the sum of theconvergence rate plus the backarc spreading. In thissetting, the subduction rate is faster than theconvergence rate, the hinge H is fixed relative to themantle, and it is retreating eastward relative to theupper plate. An example can be the Tonga subduc-tion zone (Bevis et al., 1995), where in the northernsegment the convergence rate between upper andlower plate is about 8 cm/yr, the backarc spreadingand the hinge migration amount 16 cm/yr, and theconvergence plus hinge retreat that is the subduc-tion rate should be 24 cm/yr; this is the velocity ofthe volcanic arc close to the subduction hinge andmigrating relative to a fixed Pacific plate (Bevis etal., 1995).

In the opposite model of an E- of NE-warddirected subduction, where both upper and lowerplates move relative to the mantle (Fig. 3), the lowerplate L moves westward at a rate of 10 cm/yr, whilesubducting eastward. The subduction hinge H shiftswestward at 12 cm/yr, and the upper plate U alsomoves westward at 17 cm/yr. Therefore the conver-gence rate between upper and lower plate is 7 cm/yr.In this case the shortening (5 cm/yr) is computed as

the difference in velocity between the subductionhinge migration rate and the upper plate velocity.The subduction rate would be 2 cm/yr, i.e., smallerthan the shortening. In this case the convergencerate is partitioned between the superficial shorten-ing and the subduction rate. It is worth noting thatthe subduction rate is smaller than the convergencerate. The subduction hinge retreats westwardrelative to the mantle, while it converges eastwardrelative to the upper plate. From this comparisonit is evident that at a given convergence rate, W-directed subduction zones in general transfer largerlithospheric volumes into the mantle with respect tothe opposed E- or NE-directed subduction zones.Heuret and Lallemand (2005) made an accurateworldwide analysis of the migration of the subduc-tion hinge, showing in 92% cases a motion in therange of ± 50 mm/yr.

Convergence versus Shorteningand Subduction Rates

As discussed above, the convergence rate candiffer with respect to the subduction rate, and the

FIG. 3. Absolute plate motions, e.g., relative to the mantle, show the same relationships as in the previous figure. Thecase of E- or NE-directed subduction exhibits a slower subduction rate with respect to the convergence rate. The trenchor subduction hinge H is moving west relative to the mantle, but is moving east relative to the upper plate. The greaterthe shortening in the orogen, the lower the viscosity of the upper plate. The convergence/shortening ratio is 1.4 and is afunction of the lithospheric viscosity. From this analysis, plate motions are not controlled by the subduction rate, but viceversa.

KINEMATICS OF SUBDUCTION ZONES 483

shortening rate in the orogen. We can analyze therelation between convergence and shortening. Theirratio is clearly 1 if they coincide.

Along W-directed subduction zones, the conver-gence/shortening ratio tends to be smaller than 1because the convergence rate is slower than the sub-duction rate when the subduction hinge migrateseastward relative to the upper plate, incrementingthe subduction rate, but not the upper-lower plateconvergence. The related accretionary prism peelsoff the downgoing lithosphere, generating a shorten-ing equal to the subduction amount.

The convergence rate is generally faster than theshortening rate, particularly along E- or NE-directed oceanic subduction zones, and their ratiohas to be greater than 1. In this setting, the higherthe viscosity of the continental upper plate, thelower will be the shortening in the orogen; the sub-

duction rate will also be relatively faster, but stillslower than the convergence rate. Vice versa, alower viscosity of the upper plate will allow agreater shortening in the orogen, and the subduc-tion rate will be smaller (Fig. 4). A screening of theconvergence/shortening ratio could therefore be anindirect way to infer the upper plate viscosity in thedifferent orogens. For example, in the centralAndes, the convergence is 73 mm/yr, whereas theshortening is about 40 mm/yr (Liu et al., 2000).Consequently, the convergence/shortening (C/S)ratio is about 1.8, and the subduction rate should be33 mm/yr. Computing the stress balance, the vis-cosity of the upper plate continental lithosphere inthe Andes has been inferred as low as 3 × 1021 Pa s(Husson and Ricard, 2004). The viscosity of oce-anic lithosphere tends to be generally larger, and itdecreases with depth. For example, an oceanic

FIG. 4. Four hypothetical cases of oceanic subduction, where the shortening is confined to the continental upperlithosphere and varies as a function of mantle viscosity. The greater the shortening, the smaller the subduction rate, andthe lower the viscosity (upper case). The convergence/shortening ratio can vary from 1 to ∞.

484 DOGLIONI ET AL.

lithosphere at 25 km depth might have a range ofviscosity between 1022–1027 Pa s, increasing withits age (Watts and Zhong, 2000). This may alsoexplain why the upper plate made by softer conti-nental lithosphere absorbs most of the deformation.The higher the convergence/shortening ratio (C/S),the higher the viscosity of the upper plate (Fig. 5).The minimum value of this ratio for E- or NE-directed subduction zones should be 1, where theamount of convergence equals the amount of short-ening in the belt, indicating very low viscosity, andvirtually no subduction. Subduction is inverselyproportional to the amount of shortening, and theconvergence/shortening ratio increases exponen-tially with the decrease of shortening, and theincrease of the subduction rate (Fig. 5). An attemptto compute the C/S along the Andes using REVEL(Sella et al., 2002) is shown as Figure 6, where it isdecreasing, moving southward. This signifies thatthe northern lithosphere of the Andes has a slightlyhigher viscosity. During collision, convergence rateslows down, and the shortening is propagating morepervasively into the lower plate. Therefore the C/Sdecreases during the final stages of the orogeniclife.

Therefore, along the western Pacific subductionzones, and the W-directed subduction zones ingeneral, such as the Apennines and Barbados, thesubduction rate is faster than the convergence rateinasmuch as it is augmented by hinge retreat andthe related backarc extension. Paradoxically, once

subduction is activated, the slab can continue tosink even with no convergence, or even with exten-sion between upper and lower plate; examples couldbe the Apennines, the Carpathians, and the Bandaarc. For example, the Barbados subduction hingeretreats about 2 cm/yr (Weber et al., 2001), in spiteof no apparently active convergence. A kinematicmodel of active subduction without convergence isshown in Figure 7, where the shortening is deter-mined not by an actively thrusting upper plate, butby the eastward-moving mantle, peeling off thesuperficial layers of the downgoing lower plate(Doglioni et al., 2004). On the other hand, in theeastern Pacific subduction zones, and in the E- orNE-directed subduction zones such as the Alps orHimalayas, the subduction rate is lower than theconvergence rate, but subduction does not occurwithout convergence.

Along both W- and E-/NE-directed subductionzones, the hinge migrates eastward relative to theupper plate, apart from a few exceptions like Japan.Therefore, most frequently along the W-directedsubduction zones, the hinge migrates away withrespect to the upper plate, whereas the hingemigrates toward the upper plate along E- or NE-directed subduction zones. In this interpretation,the far-field velocities of the upper and lower platescontrol the subduction rate, and subduction is a pas-sive process. In fact, rates of subduction do notdetermine plate velocities, but are rather a conse-quence of them.

FIG. 5. Diagram depicting the relations between shortening rate, convergence/shortening ratio, subduction rate, andviscosity of the upper plate along an E- or NE-directed subduction zone, at a convergence rate of 7 cm/yr.

KINEMATICS OF SUBDUCTION ZONES 485

FIG. 6. Convergence/shortening ratio along the Andes. The black line is assumed as a fixed reference in the stableupper plate, and the red line as the subduction hinge that is relatively converging, but more slowly than the Nazca–SouthAmerica convergence rate. Thin red lines are circles of the pole of the Nazca–South America motion. The higher the C/S, the higher the viscosity of the South American lithosphere.

FIG. 7. Kinematic model of a W-directed subduction where the system can be active even without convergence.Relative motions are shown in red. Shortening in the accretionary prism occurs because the mantle M moves eastwardrelative to U, peeling off the shallow layers of the lower plate. Abbreviations: U = upper plate; B = belt; L = lower plate.Modified after Doglioni et al. (2004).

486 DOGLIONI ET AL.

W- versus E-/NE-Directed Subduction and Orogens

The shortening in prisms of W-dipping subduc-tions can be larger than the convergence rate,whereas it is in general smaller than the conver-gence in orogens of E- or NE-related subductions.This occurs because along W-directed slabs theshortening is confined to the lower plate and directlyequivalent to the amount of subduction, which islarger than the convergence rate. In the E- or NE-directed slabs, the shortening is smaller than theconvergence because the convergence is partitionedpartly in the subduction rate, and partly in the con-traction of the upper plate. While W-subduction–related prisms have a forebelt and a backarc basin,E-NE-subduction–related orogens have a forebeltand a widely developed retrobelt dating from thevery early stages.

The opposite kinematics between the oppositesubduction zones can be explained by the westwarddrift of the lithosphere detected in the hotspot refer-ence frame (Gripp and Gordon, 2002), generatingdifferent geodynamic settings as a function of thepolarity of the subduction (Doglioni, 1993). The net“westward” rotation of the lithosphere provides twomain tectonic consequences: (1) the mantle is rela-tively shifting “eastward” along a sinusoidal flow;and (2) the decollement planes at the base andwithin the lithosphere are polarized toward the“west” (Doglioni et al., 1999). The result of thisasymmetry is that along W-directed subductionzones, only the shallow rocks on the subductionhinge are accreted because the accretionary prismbasal decollement is traveling at the top of thesubduction, without an actively thrusting upperplate. In the opposite E- or NE-directed subductionhinge, the lithosphere basal decollement is rampingupwards, providing a mechanism to uplift deep-seated rocks (Doglioni, 1992).

As a consequence, the shortening is concen-trated in the shallow layers of the lower plate alongW-directed subduction zones, both during theoceanic and the continental stages of subduction(Fig. 8), providing small volumes to the orogen andmaintaining a low elevation of the critical taper(e.g., Bigi et al, 2003; Lenci et al., 2004).

In contrast, shortening is mostly concentrated inthe upper continental lithosphere in E- or NE-directed oceanic subduction zones (Fig. 9). Andean-type trenches are in fact often characterized bytectonic erosion (e.g., von Huene and Lallemand,

1990; Ranero and von Huene, 2000), rather thanaccretion. However, accretion is documented alongthe Cascadia, Indonesia, and several other segmentsof this type of subduction zone (e.g., Hyndman et al.,1993), thus providing a transfer of shallow rocksfrom the lower to the upper plate. However, it is onlyat a collisional stage that the lower plate is eventu-ally entirely involved and shortened. During thisfinal stage, the lower plate contributes significantlyto the volume increase of the orogen because shearzones and decollement planes enter much deeperinto the continental crust of the footwall plate (Fig. 9).

A number of exceptions occur to these two endmembers. For example, Japan has low elevation andother characteristics of W-directed subductionzones, such as the backarc basin and an accretion-ary prism mostly composed of shallow rocks (vonHuene, 1986). However, the slab has low dip of thesubduction plane (Jarrard, 1986) and there is noactive hanging wall extension, despite an onshoremorphology suggesting widely distributed structuraldepressions. GPS data confirm that the Japan Seabackarc is not opening anymore, but rather isclosing (Mazzotti et al., 2001). This indicates thatthe system possibly has arrived at an end, and itsinversion started. Unlike other living W-directedsubduction zones, the slab hinge is moving towardthe upper plate (Fig. 10).

W-directed subduction zones on Earth appear tobe short lived. If the Marianas or Japan subductionswere active since the Cretaceous with a steady-stateeastward hinge retreat, they should be now in themiddle of the Pacific. Backarc spreading in thehanging wall of W-directed subduction zones aremostly Cenozoic, pointing out that the related sub-duction should have a similar age. Backarc basinsprobably arrive to a critical opening stage, and thenthey become closed, until a new subduction starts.The non-standard Japan subduction in fact showsscattered intermediate deep seismicity, unlike whattypically occurs along similar slabs (e.g., Omori etal., 2004).

Another apparent discrepancy to the W versus E/NE asymmetry is the occurrence of backarc basinsalso in the hanging wall of E- or NE-directedsubduction zones. However, these types of riftsrarely arrive to oceanic spreading, as W-directedsubduction do in their hanging wall. Moreover, theymay be post-subduction (e.g., Basin and Range;Doglioni, 1995), or accommodating differentialadvancement of the upper plate over the lower plate(Fig. 11). Examples are the Aegean and the Andaman

KINEMATICS OF SUBDUCTION ZONES 487

FIG. 8.W-directed subduction zones have in general the subduction hinge migrating away from the upper plate U,here considered fixed. The retreat of the slab hinge allows extension in the hanging wall and the opening of a backarcbasin. The accretionary prism is mainly composed of shallow rocks offscraped from the top of the downgoing lower plate,apart from basement slices of inherited orogens to the west. The sum of convergence and hinge retreat determines thesubduction rate. The arrival of continental lithosphere (green) at the trench slows down the subduction, but the continen-tal lithosphere can subduct at least to 150–200 km. The involvement of the continental crust allows a deeper decollementdepth of the accretionary prism, generating a thicker belt. This setting might remain active even when the convergencerate is null. In this case, the subduction rate will be the same as the hinge retreat.

488 DOGLIONI ET AL.

rifts. The extension in western Turkey, Aegean Sea,Greece, and Bulgaria can be interpreted as a resultof the differential convergence rates between theNE-directed subduction of Africa relative to thehanging wall–disrupted Eurasian lithosphere(Doglioni et al., 2002). Considering fixed Africa, thefaster southwestward motion of Greece relative toCyprus-Anatolia determines the Aegean extension.The differences in velocity can be ascribed to differ-ential decoupling with the asthenosphere. Unlikewest Pacific backarc basins where the astheno-sphere replaced a subducting and retreating slab,

the Aegean rift represents a different type of exten-sion associated with a subduction zone, where thehanging wall plate overrode the slab at differentvelocities, implying internal deformation. While W-directed subductions occur with the rollback of thelower plate relative to the upper plate, the Aegeansetting needs three plates—i.e., a common lowerplate, and two plates overriding at different veloci-ties. Analogously, assuming only few mm/yr of rela-tive motion between the Indian and Australianplates (e.g., Bird, 2003), along the Himalayas colli-sion, the Indo-Australian plates converge relative to

FIG. 9. E- or NE-directed subduction zones have typically faster convergence rates when oceanic lithosphere (black)is subducting. During oceanic subduction (upper panel), the convergence is partitioned both in the upper plate shorten-ing and in the subduction rate. The subduction hinge H moves toward the relatively fixed upper plate, and its velocityrepresents the upper-plate shortening. The orogen in the upper plate is double vergent, and no backarc basin forms. Atthe collisional stage (lower panel), the lower plate is also extensively shortened, and the western deformation border (H)is shifted westward. The convergence rate is partitioned between upper and lower plate shortening, and the subductionrate.

KINEMATICS OF SUBDUCTION ZONES 489

Asia at about 36 mm/yr, while the same Indo-Aus-tralian plates along their oceanic northern part areoverridden by the Sumatra-Burma plate at around64 mm/yr (Sella et al., 2002). Therefore, assuming arelatively coherent lower plate, the hanging wallplates move SSW-ward at different velocity, thisgradient being responsible for the extensionbetween Asia and Sumatra-Burma, and generatingthe Andaman rift. In this type of geodynamic setting,the subduction hinge still moves toward the upperplate, as in the normal E- or NE-directed subduc-tion zones.

Upper Mantle Circulation

Along oceanic ridges, melting produces a resid-ual depleted asthenosphere that is also more viscous

than the undepleted one. Braun et al. (2000) haveshown that water extraction during melting deter-mines higher viscosities up to 2 orders of magnitudein the residual mantle. The mantle, once depletedduring the transit beneath the ridge, should becooler, less dense, and more viscous. The most shal-low, depleted asthenosphere will eventually becomelithospheric mantle with the progressive cooling,shifting away from the ridge. In this view, movinghorizontally within the asthenosphere, the viscositycould be due more to lateral compositional varia-tions than to temperature gradients.

The Pacific plate is the fastest “westward”-mov-ing plate and it has the lowest viscosity of the under-lying asthenosphere (Pollitz et al., 1998). Thereforethere seems to be a positive correlation betweenasthenospheric viscosity and plate velocities.

FIG. 10. A W-directed subduction zone is generally short lived (30–50 m.y.). In the normal state (upper panel), thebackarc is opening. When the backarc stop widening, the subduction hinge moves back, inverting the basin (lowerpanel). The westward motion of the subduction hinge diminishes the subduction rate, and possibly its slab dip. Thiscould be the present kinematics of the Japan subduction zone. Velocities are relative to the mantle.

490 DOGLIONI ET AL.

The question of hotspot source depth, i.e., shal-low versus deep mantle, is under debate (Foulger etal., 2005). In the event that they are shallow, allplates move “westward” relative to the hotspot refer-ence frame because the net rotation rises to about 9cm/yr (Doglioni et al., 2005). Plate-variable veloci-ties appear to be controlled by the lithosphere-asthenosphere decoupling, which is a function ofasthenospheric viscosity. Subduction zones andoceanic ridges contribute to upper mantle convec-tion. The W-directed subductions clearly reach thelower boundary of the upper mantle. The E- or NE-directed subduction zones are rather shallower andless inclined. Most of their seismicity terminates inthe asthenosphere, except for a few deep earth-quakes at the upper-lower mantle transition. Most ofthe Earth’s subduction-related magmatism issourced from a top slab depth range of 65–130 km(England et al., 2004). As shown in the previoussection, the Andean subduction rate should beslower than the convergence rate. Moreover, the slabalong some segments of the Andes becomes almostflat, indicating the insertion of the plate into theasthenosphere. At that depth, the slab shares thepressure and temperature that allow partial meltingof the asthenosphere. Therefore the oceanic litho-sphere might re-enter the source mantle, becomingductile, and gradually melting, refertilizing themantle that generated it (Fig. 12). Note that in themantle reference frame, assuming shallow hotspots

and the faster westward drift of the lithosphere, theNazca plate is coming out of the mantle, but subduc-tion occurs because the South America plate moveswestward faster than Nazca.

W-directed subduction zones are penetratingand geochemically modifying the lower upper man-tle. E- or NE-directed subduction zones are moresuitable for regenerating the shallow upper mantlethat is the most favorable candidate for sourcingMORB along mid-oceanic ridges (e.g., Bonatti et al.,2003).

Discussion and Concluding Remarks

Along W-directed slabs, the rate of subduction isfaster than the convergence rate, whereas along E-or NE-directed slabs, the rate of subduction isslower than the convergence rate. The two oppositekinematics predict a subduction hinge migratingaway from (W-inclined subduction), or toward (E-NE-dipping subduction) the upper plate, apart froma few exceptions. The kinematics described abovesuggest that subduction zones are passive featuresrelative to far-field plate velocities, inasmuch assubduction rates can be even smaller than relativeplate motions along E- or NE-directed subductionzones. The convergence/shortening ratio is regularlysmaller than 1 in W-directed subduction zones.Because the shortening is concentrated in the accre-tionary prism consisting of the shallow crust of the

FIG. 11. Along E- or NE-directed subduction zones, the upper plate can have variable convergence rates with respectto the lower plate, being divided into two subplates (U1 and U2), and variable hinge velocity (H1 and H2). Extension canoccur in the hanging wall for these velocity gradients. Examples could be the Aegean Sea or the Andaman Sea.

KINEMATICS OF SUBDUCTION ZONES 491

lower plate (Fig. 8), and the shortening equals thesubduction, the convergence/shortening ratio isopposed and smaller with respect to the E- NE-directed subduction zones.

The convergence/shortening ratio is regularlyhigher than 1 in E- and NE-directed subductionzones (Fig. 4). This value is sensitive to the viscosityof the upper continental lithosphere. Higher ratiosmean higher viscosities of the lithosphere, i.e., it isstiffer and sustains the convergence, while most ofthe convergence is absorbed by subduction (Fig. 5).Lower ratios indicate larger values of upper-plateshortening, faster motions of the subduction hingetoward the upper plate, and lower subduction rates.During the earlier oceanic subduction, the relatedorogen is composed of upper-plate doubly-vergentshortening, whereas only at the collisional stagedoes the lower plate more widely contribute to theorogenic growth (Fig. 9). The few exceptions can beexplained either by inversion and closure of anearlier W-directed subduction-related backarcbasin, as in Japan (Fig. 10), or upper-plate variableW- or SW-ward velocity generating a sort of backarcbasin in the hanging wall of E- or NE-directed sub-duction zones, as in the New-Hebrides, Hellenic,and Western Indonesia subduction zones (Fig. 11).

W-directed subduction zones provide largervolumes of lithospheric return to the mantle than theopposite subduction zones that may instead refertil-ize the shallow upper mantle (Fig. 12). The “west-ward” drift of the lithosphere seems to be the mostfavorable mechanism for the aforementioned tuning.Moreover, the observation that the convergence dif-fers both negatively and positively from shorteningsupports the notion that plate boundaries (subduc-

tion and related orogen) are passive features, and donot provide the first-order driving energy of platemotions.

Acknowledgments

Discussions with D. Anderson, E. Bonatti, G. V.Dal Piaz, C. Faccenna, G. Foulger, F. Innocenti, J.Van Hunen, S. Lallemand, F. Mongelli, L. Roydenand S. Willett were greatly appreciated.

REFERENCES

Ampferer, O., 1906, Über das Bewegungbild von Falt-engebirge, Austria: Jahrbuch der Kaiserlich-Königli-chen Geologischen Reichsanstalt, Wien, v. 56, p. 539–622.

Anderson, D. L., 1989, Theory of the Earth: Oxford, UK,Blackwell, 366 p.

Anderson, D. L., 2001, Topside tectonics: Science, v. 293,p. 2016–2018.

Bevis, M., Taylor, F. W., Schutz, B. E., Recy, J., Isacks,B. L., Helu, S., Singh, R., Kendrick, E., Stowell, J.,Taylor, B., and Calmant, S., 1995, Geodetic observa-tions of very rapid convergence and back-arc exten-sion at the Tonga arc: Nature, v. 374, p. 249–251.

Bigi, S., Lenci, F., Doglioni, C., Moore, J. C., Carminati,E., and Scrocca, D., 2003, Decollement depth vs.accretionary prism dimension in the Apennines andthe Barbados : Tectonics , v. 22, 2 , p. 1010[doi:10.1029/2002TC001410].

Bird, P., 2003, An updated digital model of plate bound-aries: Geochemisty, Geophysics, Geosystems, v. 4, no.3, p. 1027 [doi:10.1029/2001GC000252].

Bonatti, E., Ligi, M., Brunelli, D., Cipriani, A., Fabretti,P., Ferrante, V., and Ottolini, L., 2003, Mantle thermal

FIG. 12. Model for the upper mantle cycle. The lower the asthenospheric viscosity, the faster the westward displace-ment of the overlying plate. The asthenospheric depletion at oceanic ridges makes the layer more viscous and decreasesthe lithosphere/asthenospheric decoupling, and the plate on the east is then slower. The oceanic lithosphere subductingeastward enters the asthenosphere, where it is molten again, to refertilize the asthenosphere. W-directed subductionsprovide deeper circulation. Note that the E-directed subduction (the Andes) tends to escape from the mantle, but it isoverridden by the upper plate (South America).

492 DOGLIONI ET AL.

pulses below the Mid-Atlantic Ridge and temporalvariations in the oceanic lithosphere: Nature, v. 423, p.499–505.

Bostrom, R.C., 1971, Westward displacement of the litho-sphere: Nature, v. 234, p. 356–538.

Braun, M. G., Hirth, G., and Parmentier, E. M., 2000, Theeffects of deep damp melting on mantle flow and meltgeneration beneath mid-ocean ridges: Earth and Plan-etary Science Letters, v. 176, p. 339–356.

Carminati, E., Doglioni, C., and Scrocca, D., 2004, Alpsvs. Apennines, in Special volume of the ItalianGeological Society for the IGC 32 Florence–2004, p.141–151.

Conrad, C. P., and Lithgow-Bertelloni, C., 2003, Howmantle slabs drive plate tectonics: Science, v. 298, p.207–209.

Doglioni, C., 1992, Main differences between thrust belts:Terra Nova, v. 4, p. 152–164.

Doglioni, C., 1993, Geological evidence for a globaltectonic polarity: Journal of the Geological Society ofLondon, v. 150, p. 991–1002.

Doglioni, C., 1994, Foredeeps versus subduction zones:Geology, v. 22, p. 271–274.

Doglioni, C., 1995, Geological remarks on the relation-ships between extension and convergent geodynamicsettings: Tectonophysics, v. 252, p. 253–267.

Doglioni, C., Agostini, S., Crespi, M., Innocenti, F.,Manetti, P., Riguzzi, F., and Savascin, Y., 2002, On theextension in western Anatolia and the Aegean sea:Journal of the Virtual Explorer, v. 7, p. 117–131.

Doglioni, C., Green, D., and Mongelli, F., 2005, On theshallow origin of hotspots and the westward drift of thelithosphere, in Foulger, G. R., Natland, J. H., Presnall,D. C., and Anderson, D. L., eds., Plates, plumes, andparadigms: Geological Society of America, SpecialPaper, v. 388, p. 735–749.

Doglioni, C., Harabaglia, P., Merlini, S., Mongelli, F.,Peccerillo, A., and Piromallo, C., 1999, Orogens andslabs vs. their direction of subduction: Earth ScienceReviews, v. 45, p. 167–208.

Doglioni, C., Innocenti, F., Morellato, C., Procaccianti, D.,and Scrocca, D., 2004, On the Tyrrhenian sea opening:Memorie Descrittive della Carta Geologica d’Italia, v.64, p. 147–164.

England, P., Engdahl, R., and Thatcher W., 2004, System-atic variation in the depths of slabs beneath arc volca-noes: Geophysical Journal International, v. 156, p.377–408.

Ernst, W. G., 2005, Alpine and Pacific styles of Phanero-zoic mountain building: Subduction-zone petrogenesisof continental crust: Terra Nova, v. 17, p. 165–188[doi: 10.1111/j.1365-3121.2005.00604.x].

Foulger, G. R., Natland, J. H., Presnall, D. C., and Ander-son, D. L., 2005, Plates, plumes and paradigms:Geoogical. Society of America, Special Paper, v. 388.

Garfunkel, Z., Anderson, C. A., and Schubert, G., 1986,Mantle circulation and the lateral migration of sub-

ducted slabs: Journal of Geophysical Research v. 91, p.7205–7223.

Gripp, A. E., and Gordon, R. G., 2002, Young tracks ofhotspots and current plate Velocities: GeophysicalJournal International, v. 150, p. 321–361.

Hacker, B. R., Abers, G. A., and Peacock, S. M., 2003,Subduction factory 1: Theoretical mineralogy, densi-ties, seismic wave speeds and H2O contents: Journal ofGeophysical Research, v. 108, B1, p. 2029[doi:10.1029/2001JB001127].

Hermann, J., 2002, Experimental constraints on phaserelations in subducted continental crust: Contributionsto Mineralogy and Petrology, v. 143, p. 219–235.

Heuret, A., and Lallemand, S., 2005, Plate motions, slabdynamics, and back-arc deformation: Physics of theEarth and Planetary Interiors, v. 149, p. 31–51.

Husson, L., and Ricard, Y., 2004, Stress balance abovesubduction: Application to the Andes: Earth and Plan-etary Science Letters, v. 222, p. 1037–1050.

Hyndman, R. D., Wang, K., Yuan, T., and Spence, G. D.,1993, Tectonic sediment thickening, fluid expulsion,and the thermal regime of subduction zone accretion-ary prisms: The Cascadia margin off Vancouver Island:Journal of Geophysical Research, v. 98, p. 21,865–21,876.

Isacks, B., and Molnar, P., 1971, Distribution of stresses inthe descending lithosphere from a global survey offocal-mechanism solutions of mantle earthquakes:Reviews of Geophysics, v. 9, p. 103–174.

Jarrard, R. D., 1986, Relations among subduction param-eters: Reviews of Geophysics, v. 24, p. 217–284.

Lallemand, S., Heuret, A., and Boutelier, D., 2005, On therelationships between slab dip, back-arc stress, upperplate absolute motion, and crustal nature in subduc-tion zones: Geochemisty, Geophysics, Geosystems, v. 6[doi:10.1029/2005GC000917].

Lenci, F., Carminati, E., Doglioni, C., and Scrocca, D.,2004, Basal décollement and subduction depth vs.topography in the Apennines-Calabrian arc: Bollettinodella Società Geologica Italiana, v. 123, p. 497–502.

Liu, M., Yang, Y., Stein, S., Zhu, Y., and Engeln, J., 2000,Crustal shortening in the Andes: Why do GPS ratesdiffer from geological rates?: Geophysical ResearchLetters, v. 27, no. 18, p. 3005–3008.

Mariotti, G., and Doglioni, C., 2000, The dip of the fore-land monocline in the Alps and Apennines: Earth andPlanetary Science Letters, v. 181, p. 191–202.

Mazzotti, S., Henry, P., and Le Pichon, X., 2001, Transientand permanent deformation of central Japan estimatedby GPS 2. Strain partitioning and arc-arc collision:Earth and Planetary Science Letters, v. 184, p. 455–469.

Mitchell, A. H. G., and Garson, M. S., 1981, Mineraldeposits and global tectonic settings: London, UK,Academic Press.

Nishiwaki, C., and Uyeda, S., 1983, Accretion tectonicsand metallogenesis, in Hashimoto, M., and Uyeda, S.,

KINEMATICS OF SUBDUCTION ZONES 493

eds., Accretion tectonics in the Circum-PacificRegions: Proceedings of the Oji International Seminaron Accretion Tectonics, Japan, 1981: Tokyo, Japan,Terra Scientific Publishing Company, Advances inEarth and Planetary Sciences, p. 349–355.

Niu, Y., O’Hara, M. J., and Pearce, J. A., 2003, Initiationof subduction zones as a consequence of lateral com-positional buoyancy contrast within the lithosphere: Apetrological perspective: Journal of Petrology, v. 44, p.851–866.

Omori, S., Komabayashi, T., and Maruyama, S., 2004,Dehydration and earthquakes in the subducting slab:Empirical link in intermediate and deep seismiczones: Physics of the Earth and Planetary Interiors, v.146, p. 297–311.

Pollitz, F. F., Bürgmann, R., and Romanowicz, B., 1998,Viscosity of oceanic asthenosphere inferred fromremote triggering of earthquakes: Science, v. 280, p.1245–1249.

Ranalli, G., Pellegrini, R., and D’Offizi, S., 2000, Timedependence of negative buoyancy and the subductionof continental lithosphere: Journal of Geodynamics, v.30, no. 5, p. 539–555.

Ranero, C., and von Huene, R., 2000, Subduction erosionalong the Middle America convergent margin: Nature,v. 344, p. 31–36.

Royden, L. H., and Burchfiel, B. C., 1989, Are systematicvariations in thrust belt style related to plate boundaryprocesses? (The Western Alps versus the Carpathians):Tectonics, v. 8, p. 51–62.

Scoppola, B., Boccaletti, D., Bevis, M., Carminati, E., andDoglioni, C., 2006, The westward drift of the litho-sphere: A rotational drag?: Bulletin of the Geological

Society of America, v. 118, Nos. 1–2, p. 199–209[doi:10.1130/B25734.1].

Sella, G. F., Dixon, T. H., and Mao, A., 2002, REVEL: Amodel for recent plate velocities from space geodesy:Journal of Geophysical Research, v. 107, B4, p.10.1029/2000JB000033.

Spence, W., 1987, Slab pull and the seismotectonics ofsubducting lithosphere: Reviews of Geophysics, v. 25,p. 55–69.

Stern, R. J., 2002, Subduction zones: Reviews of Geophys-ics, v. 40, p. 1012 [doi:10.1029/2001RG000108].

Tatsumi, Y., and Eggins S., 1995, Subduction zone mag-matism: Oxford, UK, Blackwell Scientific, Frontiers inEarth Sciences, 211 p.

von Huene, R., 1986, Seismic images of modern conver-gent margin tectonic structure: Bulletin of the Ameri-can Association of Petroleum Geologists, v. 26, p. 1–60.

von Huene, R., and Lallemand, S., 1990, Tectonic erosionalong the Japan and Peru convergent margins: Geolog-ical Society of America Bulletin, v. 102, p. 704–720.

Waschbusch, P., and Beaumont, C., 1996, Effect of slabretreat on crustal deformation in simple regions ofplate convergence: Journal of Geophysical Research,v. 101, B12, p. 28,133–28,148.

Watts, A. B., and Zhong S., 2000, Observations of flexureand rheology of oceanic lithosphere: Geophysical Jour-nal International, v. 142, p. 855–875.

Weber, J. C., Dixon, T. H., DeMets, C., Ambeh, W. B.,Jansma, P., Mattioli, G., Saleh, J., Sella, G., Bilham,R., and Pérez, O., 2001, GPS estimate of relativemotion between the Caribbean and South Americanplates, and geologic implications for Trinidad andVenezuela: Geology, v. 29, p. 75–78.